We keep getting a difference between the two ways at the second level.
I have attached a sample subject 2nd level zstats map with
histograms. The red-yellow part shows what is common between the 1
subject model and the all subjects model. The blue part shows the
extra activation from the 1 subject model which is mostly outside the
brain. The top histogram is for the all subjects model and the lower
one is the one subject model. Any ideas as to why the dip occurs from
-2.5 to -2 on the lower histogram? This seems to occur in a
consistent manner for all subjects we have analysed.
(Sorry about the image quality but it has be under the 50k limit for
FSL postings)
Thank you,
Marc
On Mon, Apr 19, 2010 at 2:15 PM, Stephen Smith <
[log in to unmask]> wrote:
Hi, sorry that wording is a little confusing in the practical web page - a
hangover from before we started recommending using FE. When you use FE
the session-session variance is IGNORED, it's really just a pooling of the
lower-level results, so that's why we're saying that it doesn't make a big
difference which way the second-level is done in this case.
Cheers.
On 19 Apr 2010, at 14:29, Marc Bouffard wrote:
Thank you for your reply. But it says in:
http://www.fmrib.ox.ac.uk/fslcourse/lectures/practicals/feat3/index.htm
under:
Group difference with multiple sessions for each subject
that we should put all the second-level analyses into a single
second-level model when there is a small number of runs for each
subject because if we use one model/subject it would not lead to a
good within subject session-to-session variance. So isn't that a
strong reason to chose the single second-level model?
Thanks again,
Marc
On Sun, Apr 18, 2010 at 12:21 PM, Stephen Smith <[log in to unmask]>
wrote:
Hi - yes, the two are not exactly the same, but I don't think there is a
strong reason for choosing one versus the other.
Cheers.
On 14 Apr 2010, at 21:21, Marc Bouffard wrote:
Hello,
I think this has been covered in the forums before but I have not found a
clear answer. What is the recommended way to combine/average fmri runs
within subjects (following first-level analysis)? For example, if for each
subject we have 2 fmri runs should we use the method outlined in one
analysis/one model:
http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html#MultiSessionMultiSubject
or should we have one separate model/analysis for each subject?
I have tried both and get simlar results but the zstats from the latter are
noisier particularly on the edges of the zstats images. So this suggests
that the two are not doing exactly the same computations and therefore are
not answering the averaging question in exactly the same way?
Regards,
Marc
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
<2nd_level_test.rar>